Is Your Company Ready for AI? Five Plain Questions to Answer First
I'm Ty Wells, founder of Rocksteady. I'm a fractional AI advisor: I get organizations ready for AI, then build what sticks.
Most AI projects in small and mid-sized organizations don't fail because the tool is bad. They fail because the business wasn't ready for it. The problem was fuzzy, the data was a mess, nobody owned it, or nobody could say whether it worked.
You can find most of that out before you spend a dollar. Here are five plain questions. Answer them honestly, in writing, with the people who actually do the work in the room.
1. What exact problem are you trying to fix?
"We should be using AI" is not a problem. It's a feeling. Start with a task instead.
Run this check:
- Name one task. Not a department, a task. "Answering the same ten customer questions by email" is a task. "Customer service" is not.
- Write down who does it today, how often, and roughly how long it takes. Use your own rough numbers. A guess you can check later beats no number.
- Write down what goes wrong. Slow? Inconsistent? Error-prone? Boring work burning out good people?
- Ask: if this task got twice as fast tomorrow, would anyone notice? If the answer is no, pick a different task.
Watch for: Starting with a tool you saw in a demo and then hunting for a problem it fits. Work backward from the pain, not forward from the product.
2. Is your information where you think it is, and is it usable?
AI works on what you give it. If your information is scattered, outdated, or locked in someone's head, the AI will be too.
Run this check:
- For the task you picked, list every place the information lives. Shared drives, inboxes, a CRM, spreadsheets, paper, one person's memory.
- Open three real examples. Are they complete? Current? Consistent with each other?
- Ask who can access each source today, and whether they should.
- Find the "source of truth." If two documents disagree, which one wins? If nobody knows, that's your first job, and it has nothing to do with AI.
Watch for: The policy doc from three years ago that everyone ignores but nobody deleted. An AI tool will happily treat it as current. Clean up or archive before you connect anything.
3. Do you know your rules on data, privacy and vendors?
This is the question people skip, and it's the one that causes the real trouble.
Run this check:
- Classify the information involved in your task. Is any of it customer personal data, employee records, health, financial, or legal information? Anything under a contract with confidentiality terms?
- Find out what laws and industry rules apply to that data where you operate and where your customers are. These rules change and differ by location and industry. Confirm current requirements with the official regulator or a qualified advisor, not a blog post. That includes this one.
- Read the terms of any AI tool you're considering. Look specifically for: whether your inputs are used to train their models, where data is stored, how long it's kept, and how you delete it. Vendors update these terms, so check the current version on the vendor's official site and note the date you read it.
- Check what your staff are already doing. Ask plainly, without blame: "Are you pasting work into any AI tools right now?" Many are. You'd rather know.
Watch for: No written policy at all. Even one page that says what can and can't go into AI tools is better than silence. Silence means everyone makes their own rules.
4. Who owns it, and do they have the time?
Every tool that sticks has an owner. Every tool that dies quietly didn't.
Run this check:
- Name one person who will own this. Not a committee. One name.
- Ask that person what they'll stop doing to make room. If the answer is "nothing," they don't have the time, and the project will slide.
- Identify the people whose daily work changes. Talk to them before you pick anything. They know where the task actually breaks, and they'll decide whether the new way gets used.
- Decide who has final say when the AI output is wrong. Someone has to check the work, especially early on.
Watch for: The leader who's excited and the team that's quietly worried about their jobs. Address it directly. Tell people what's changing, what isn't, and why. Fear kills adoption faster than bad software.
5. How will you know it worked?
If you can't say what success looks like before you start, you won't be able to tell afterward either.
Run this check:
- Go back to the numbers from question one. Pick one or two to track: time per task, errors caught, response time, rework.
- Measure them now, before anything changes. A week of honest tracking is enough to have a baseline.
- Set a short trial window and a date to review it. Put the review on the calendar now.
- Decide in advance what result means "keep going," "adjust," and "stop." Write it down so the decision isn't made on vibes later.
- Make sure you can stop. Know how to export your data and cancel before you sign up.
Watch for: Measuring usage instead of results. "Everyone logged in" tells you nothing. "The task takes less time and has fewer mistakes" tells you something.
What your answers are telling you
Lay your five answers side by side.
If you have clear answers to all five, you're in good shape. Start small with the one task you picked, run the trial, and review on the date you set.
If you got stuck on one or two, that's normal. Fix those first. Most of the fixes are cheap: a cleanup session on your files, a one-page data policy, a conversation with the team, a week of tracking. None of them need AI.
If you got stuck on three or more, you're not ready yet, and that's useful to know. Spending money on tools now would mostly buy frustration. The readiness work is the project for now.
If you want a second set of eyes on your answers, a first conversation with Rocksteady is a free 15-minute fit call booked at rocksteady.ai.
A few habits that help either way
- Pick one task, not five. One win you can measure teaches you more than five half-finished pilots.
- Keep a human checking the output until you've seen enough of it to trust it, and keep spot-checking after.
- Write things down. The problem, the baseline, the rules, the owner, the review date. A one-page document is enough.
- Revisit your policy and vendor terms on a schedule. Laws and vendor terms change. Set a reminder to recheck them against the official sources.
- Talk to the people doing the work. Every time. They'll tell you what the dashboard won't.
Readiness isn't glamorous. It's mostly cleanup, decisions and conversations. But it's the difference between AI that gets used and AI that gets quietly abandoned.